Epidural anesthesia needle guidance by forward-view endoscopic optical coherence tomography and deep learning

Area of Science:

• Medical Imaging
• Spinal Procedures
• Artificial Intelligence in Medicine

Background:

• Accurate epidural needle placement is critical for effective anesthesia and patient safety.
• Current methods for guiding epidural needle insertion face challenges in real-time tissue visualization.
• Identifying anatomical layers and maintaining safe distances during puncture is complex.

Purpose of the Study:

• To develop and validate a forward-view endoscopic optical coherence tomography (OCT) system for real-time imaging during epidural needle insertion.
• To create deep learning models for automated tissue layer recognition and distance measurement ahead of the needle tip.
• To assess the technical feasibility of this integrated imaging and AI approach for improving epidural anesthesia procedures.

Main Methods:

• Development of a forward-view endoscopic OCT system for in-situ tissue imaging.
• Application of deep learning, specifically binary classification models (Inception architecture), for analyzing OCT data.
• Testing the system in porcine spine models to evaluate tissue layer identification and distance estimation accuracy.

Main Results:

• Deep learning models achieved an average classification accuracy of 96.65% for identifying five distinct spinal tissue layers.
• Regression models accurately estimated the distance to the dura mater with a mean absolute percentage error of 3.05% ± 0.55%.
• The OCT system demonstrated real-time imaging capabilities for guiding needle placement.

Conclusions:

• The developed endoscopic OCT system combined with deep learning models is technically feasible for real-time tissue recognition during epidural needle placement.
• This novel imaging strategy can automatically identify spinal structures and measure critical distances, potentially improving procedural safety and accuracy.
• The findings support the integration of advanced imaging and AI for enhanced minimally invasive spinal interventions.